$f$-FUM: Federated Unlearning via min--max and $f$-divergence
Radmehr Karimian, Amirhossein Bagheri, Meghdad Kurmanji, Nicholas D. Lane, Gholamali Aminian
TL;DR
f-FUM formulates federated unlearning as a min–max optimization using $f$-divergences to maximize discrepancy on forgotten data while minimizing degradation on retained data. The framework operates as a plugin to existing FL setups, with a gradient-based server–client protocol and post-training refinements, accommodating both robustness- and privacy-oriented unlearning. Empirical results across CIFAR-10, Fashion-MNIST, and MNIST show that JS- and KL-based divergences yield robust forgetting with favorable utility and privacy trade-offs, often outperforming baselines. The work highlights a practical, scalable approach to data deletion and poisoning mitigation in decentralized learning with minimal retraining. It also discusses the privacy-utility trade-offs inherent to distributed deletion requests and points to future directions in extending the theoretical guarantees and exploring additional divergences.
Abstract
Federated Learning (FL) has emerged as a powerful paradigm for collaborative machine learning across decentralized data sources, preserving privacy by keeping data local. However, increasing legal and ethical demands, such as the "right to be forgotten", and the need to mitigate data poisoning attacks have underscored the urgent necessity for principled data unlearning in FL. Unlike centralized settings, the distributed nature of FL complicates the removal of individual data contributions. In this paper, we propose a novel federated unlearning framework formulated as a min-max optimization problem, where the objective is to maximize an $f$-divergence between the model trained with all data and the model retrained without specific data points, while minimizing the degradation on retained data. Our framework could act like a plugin and be added to almost any federated setup, unlike SOTA methods like (\cite{10269017} which requires model degradation in server, or \cite{khalil2025notfederatedunlearningweight} which requires to involve model architecture and model weights). This formulation allows for efficient approximation of data removal effects in a federated setting. We provide empirical evaluations to show that our method achieves significant speedups over naive retraining, with minimal impact on utility.